Functional Losses in Ground Spider Communities Due to Habitat Structure Degradation Under Tropical Land-Use Change

Ecology, 101(3), 2020, e02957 © 2019 by the Ecological Society of America

Functional losses in ground spider communities due to habitat structure degradation under tropical land-use change

1,2,13 3 4,5,6 3 1 ANTON M. POTAPOV , NADINE DUPERR E, MALTE JOCHUM , KERSTIN DRECZKO, BERNHARD KLARNER, 5,7 1 8,9 9,10 ANDREW D. BARNES , VALENTYNA KRASHEVSKA , KATJA REMBOLD, HOLGER KREFT , 5,11 12 3 1,10 ULRICH BROSE, RAHAYU WIDYASTUTI, DANILO HARMS, AND STEFAN SCHEU 1J.F. Blumenbach Institute of Zoology and Anthropology, University of Gottingen,€ Untere Karspule€ 2, 37073 Gottingen,€ Germany 2A.N. Severtsov Institute of Ecology and Evolution, Russian Academy of Sciences, Leninsky Prospect 33, 119071 Moscow, Russia 3Center of Natural History, Zoological Museum, Universitat€ Hamburg, Bundesstraße 52, 20146 Hamburg, Germany 4Institute of Plant Sciences, University of Bern, Hochschulstrasse 6, 3012 Bern, Switzerland 5German Centre for Integrative Biodiversity Research (iDiv), Deutscher Pl. 5E, 04103 Leipzig, Germany 6Institute of Biology, Leipzig University, Deutscher Platz 5e, 04103 Leipzig, Germany 7School of Science, the University of Waikato, Private Bag 3105, 3240 Hamilton, New Zealand 8Botanical Garden of the University of Bern, Altenbergrain 21, 3013 Bern, Switzerland 9Biodiversity, Macroecology & Biogeography, University of Gottingen,€ Busgenweg€ 1, 37077 Gottingen,€ Germany 10Centre of Biodiversity and Sustainable Land Use, Von-Siebold-Strasse 8, 37075 Gottingen,€ Germany 11Institute of Biodiversity, Friedrich Schiller University, Dornburger Strasse 159, 07743 Jena, Germany 12Department of Soil Sciences and Land Resources, Institut Pertanian Bogor (IPB), Jln. Meranti Kampus IPB Darmaga, 16680 Bogor, Indonesia

Citation: Potapov, A. M., N. Duperre, M. Jochum, K. Dreczko, B. Klarner, A. D. Barnes, V. Krashevska, K. Rembold, H. Kreft, U. Brose, R. Widyastuti, D. Harms, and S. Scheu. 2020. Functional losses in ground spider communities due to habitat structure degradation under tropical land-use change. Ecology 101(3):e02957. 10.1002/ecy.2957

Abstract. Deforestation and land-use change in tropical regions result in habitat loss and extinction of species that are unable to adapt to the conditions in agricultural landscapes. If the associated loss of functional diversity is not compensated by species colonizing the converted habitats, extinctions might be followed by a reduction or loss of ecosystem functions including biological control. To date, little is known about how land-use change in the tropics alters the functional diversity of invertebrate predators and which key environmental factors may mitigate the decline in functional diversity and predation in litter and soil communities. We applied litter sieving and heat extraction to study ground spider communities and assessed structural character- istics of vegetation and parameters of litter in rainforest and agricultural land-use systems (jungle rubber, rubber, and oil palm monocultures) in a Southeast Asian hotspot of rainforest conver- sion: Sumatra, Indonesia. We found that (1) spider density, species richness, functional diversity, and community predation (energy flux to spiders) were reduced by 57–98% from rainforest to oil palm monoculture; (2) jungle rubber and rubber monoculture sustained relatively high diversity and predation in ground spiders, but small cryptic spider species strongly declined; (3) high spe- cies turnover compensated losses of some functional trait combinations, but did not compensate for the overall loss of functional diversity and predation per unit area; (4) spider diversity was related to habitat structure such as amount of litter, understory density, and understory height, while spider predation was better explained by plant diversity. Management practices that increase habitat-structural complexity and plant diversity such as mulching, reduced weeding, and intercropping monocultures with other plants may contribute to maintaining functional diversity of and predation services provided by ground invertebrate communities in plantations. Key words: Araneae; belowground biodiversity; energy flux; functional diversity; land-use change; oil palm; predation; rainforest; rubber.

regions (Gibbs et al. 2010). These changes are driven pri- INTRODUCTION marily by increasing demand in agricultural products, Global deforestation and agricultural intensification such as palm oil and rubber (Corley and Tinker 2016, are causing strong environmental changes in tropical Clough et al. 2016). Recent land-use change is particu- larly prominent in Indonesia, which has experienced the strongest oil palm expansion and is among the leaders in Manuscript received 21 April 2019; revised 31 August 2019; accepted 11 November 2019. Corresponding Editor: Elizabeth the global chart of deforestation rates (Margono et al. T. Borer. 2014, Corley and Tinker 2016, FAOSTAT 2016). Defor- 13 E-mail: [email protected] estation and agricultural expansion in tropical regions is

Article e02957; page 1 Article e02957; page 2 ANTON M. POTAPOV ET AL. Ecology, Vol. 101, No. 3 associated with a decline in taxonomic diversity (Fitzher- higher pH (Krashevska et al. 2015, Allen et al. 2016), bert et al. 2008, Drescher et al. 2016) and in the perfor- lower plant diversity, but higher density of understory mance of many ecosystems functions (Dislich et al. compared to rainforest sites (Rembold et al. 2017). 2017). Predation is among the key ecosystem functions These complex changes result in a distinctly different that decline in agricultural landscapes, leading to a habitat structure, suggesting potential turnover of spe- potential weakening of natural pest control (Tscharntke cies and functional traits in soil communities between et al. 2005, Gurr et al. 2017). The effect of tropical land- these different land-use systems (Turner and Foster use change on predation has been studied mostly for ver- 2009). It is not clear if this community change can sus- tebrate predators that control populations of large-sized tain the predation function typically provided by pris- insects, but only a few studies have focused on inverte- tine ecosystems. A number of studies found abundance brate predators that control small-sized litter and soil and species richness in litter and soil arthropods to be fauna (Barnes et al. 2014, Klarner et al. 2017, Potapov lower in oil palm plantations compared to rainforest et al. 2019). (Chung et al. 2000, Fayle et al. 2010, Barnes et al. The relationship between the loss of taxonomic 2014), but selected groups of decomposers, such as diversity and ecosystem functions can be disentangled woodlice and earthworms, may have higher density in with functional traits, i.e., characteristics of individuals monoculture plantations (Hassall et al. 2006, Turner that are relevant to their role in ecosystems and govern and Foster 2009, Potapov et al. 2019). Due to large their response to environmental changes (Vandewalle body size, however, increased earthworm populations et al. 2010, Gagic et al. 2015). Trait-based indices, such in plantations contribute little energy to higher trophic as functional diversity and functional redundancy, pro- levels of the invertebrate food web (Potapov et al. vide proxies for overall changes in ecosystem function- 2019). Such differential responses of various animal ing and resilience (Mason et al. 2005, Laliberteand taxa imply selective exclusion and/or decline of species. Legendre 2010). Land-use change may result in the loss This raises questions on which functional traits con- of species, which cannot adapt or colonize plantations tribute to the resilience against habitat change (Larsen rapidly enough and this may be followed by the loss of et al. 2005), which abiotic and biotic factors drive these functional traits and thereby the functional roles these changes and how differential responses affect the species play. The loss of ecosystem functions thus ecosystem service of predation provided by tropical depends on both the functional redundancy of the orig- invertebrate predator communities. inal community and the magnitude of species loss Predators with a high flexibility may persist under (Flynn et al. 2009). If the original community has many tropical land-use change, maintaining their abundance functionally redundant species, random species loss (e.g., centipedes; Klarner et al. 2017), but little is may initially have only little effect on functional diver- known about other major groups of predators such as sity. However, if there is a factor acting nonrandomly in spiders, despite their importance for ecosystem func- respect to functional traits (or if functional redundancy tioning (Nyffeler and Birkhofer 2017). With more than of the original community is low), functional diversity 48,000 described species (World Spider Catalog; avail- may be lost at a rate that equals or even exceeds species able online),14 spiders (Araneae) are among the most loss (Flynn et al. 2009). These different scenarios result diverse and abundant invertebrate predator taxa world- in a minor or prominent decrease in ecosystem func- wide (Wise 1995, Nyffeler and Birkhofer 2017). Differ- tioning, respectively. On the other hand, plantations ent groups of spiders are adapted to specific are colonized by new species, which replace some spe- microhabitats characterized by differences in food cies that were lost. Such species turnover may compen- availability, vegetation structure and moisture, and sate for the loss of ecosystem functions, or at least employing diverse hunting strategies to target a wide maintain the overall level of functional diversity (Aslan array of prey species (Wise 1995). Spiders span several et al. 2014, Garcıa et al. 2014). To date, very little is orders of magnitude in body mass, thus consuming known about the rate of species vs. trait loss and turn- prey of a wide size range (Herberstein 2011). The diver- over in invertebrate communities facing land-use sity and versatility of spiders make them play a signifi- change (Rigal et al. 2018). cant role in terrestrial food webs and render them Litter and soil harbor about 50% of animal biomass exceptionally suitable for analyzing changes in preda- globally (Fierer et al. 2009), which comprises diverse tion as a major ecosystem function affected by distur- invertebrate communities that form complex food webs, bance (Prieto-Benıtez and Mendez 2011, Nyffeler and contributing to nutrient cycling and stability of terres- Birkhofer 2017). trial ecosystems (Bardgett and van der Putten 2014). In Here, we investigated changes in abundance, diversity, contrast to relatively well-studied effects of tropical energy flux to spiders (as a proxy for predation), and land use on aboveground biota, only few community- species and functional trait composition in ground-asso- level studies addressed these effects in litter and soil ciated spider communities across sites in tropical rain- (Foster et al. 2011). Oil palm and rubber plantations forest (as reference), rubber agroforest (“jungle rubber”) are characterized by a reduction in the amount and nutritional quality of litter (Krashevska et al. 2015), 14 https://wsc.nmbe.ch March 2020 PREDATION LOSS IN TROPICAL PLANTATIONS Article e02957; page 3 and monocultures of rubber and oil palm (as intensively analyzed using an elemental analyzer, and amount of lit- managed agroecosystems). We applied two sampling ter was determined gravimetrically (for details see Kra- methods, litter sieving (LS) and heat extraction (HE), to shevska et al. 2015). Density of earthworms in the litter gain a comprehensive understanding of land-use-driven and in the upper 5 cm of soil was investigated in three changes in the spider communities. We focused on three subplots (a, b, c) in 2013 using heat extraction (for research questions: (1) How pronounced is the effect of details, see Potapov et al. 2019). rainforest conversion into plantations on spider density, species richness, functional diversity, and energy flux to Spider collection spiders? (2) Does land-use change differentially affect species with different functional traits and which envi- Different parts of the spider community were assessed ronmental factors drive these effects? (3) What is the rel- in two separate sampling campaigns. The first campaign ative rate of species and functional trait loss and (litter sieving, LS) was conducted in October–November turnover with land-use change? 2012 (Barnes et al. 2014). In three subplots per plot (a, b, c), the litter from a 1 9 1 m area was sieved through a 2-cm mesh sieve. All spiders visible by eye were hand- METHODS collected from the siftings and stored in 65% ethanol. The method represents large and mobile spider species. Study sites A total of 1,081 individuals were collected (Data S1: The study was conducted in the framework of the CommunityMatrix_spiders). The second campaign (heat EFForTS project (Georg-August-Universitat€ Gottingen€ extraction, HE) was conducted in October–November 2019), investigating ecological and socioeconomic 2013 (Klarner et al. 2017). On each plot, the same sub- changes associated with the transformation of lowland plots (a, b, c) were sampled with a spade. One soil sam- rainforest into agricultural systems (Drescher et al. ple of 16 9 16 cm was taken on each subplot, 2016). Four land-use systems, rainforest (forest, F), jun- comprising the litter layer and the underlying mineral gle rubber agroforest (J), rubber monoculture (R), and soil to the depth of 5 cm. Soil and litter from the second oil palm monoculture (O; Fig. 1), were studied across sampling campaign were transported in individual plas- two regions (Harapan and Bukit Duabelas landscape) in tic containers and processed separately, but data were Jambi Province, Sumatra, Indonesia. Each system was pooled for statistical analyses (but see Appendix S2). investigated in eight replicate plots, four in each of the Spiders were extracted by heat (Kempson et al. 1963), two landscapes, resulting in a total of 32 plots of collected in a dimethyleneglycol-water solution (1:1), 50 9 50 m with five 5 9 55 m subplots randomly and thereafter transferred into 70% ethanol. The method placed per plot. Forests were classified as primary represents small and cryptic spider species. A total of degraded rainforest (Margono et al. 2014), jungle rubber 331 individuals were collected (Data S1: Commu- sites originating from disturbed rainforests enriched with nityMatrix_HEspiders). rubber trees (Hevea brasiliensis); rubber and oil palm (Elaeis guineensis) monocultures were 12–14 yr old at Spider identification the time of sampling. All adult spiders (43% of all individuals) were identi- fied to (morpho)species using a Leica M125 (Leica Soil and vegetation parameters Microsystems, Wetzlar, Germany) dissecting microscope Characteristics of the litter and the vegetation were and available identification keys for spiders in Asia, i.e., assessed on each sampling site (Appendix S1: Table S1). Barrion et al. (1995), Deeleman-Reinhold (2001), and During a vegetation survey, conducted from February Jocque and Dippenaar-Schoeman (2006). In total, 139 2013 to August 2014, all trees with a diameter at breast species from 33 families were identified with 131 species height ≥10 cm within the entire 50 9 50 m plot area, found in the litter-sieving samples and 30 species in the were identified and counted. Understory plants were heat-extracted samples with an overlap of 22 species assessed in the five 5 9 5 m subplots (a, b, c, d, e) within between the methods (Appendix S2: Fig. S1). Species each plot: all vascular plant individuals growing within coverage (proportion of the collected species in the the subplots were identified and measured (height). regional gamma diversity) was estimated as 41–80% in Based on these data, we calculated tree species richness the litter-sieving samples and 60–86% in the heat-ex- (TreeRich) and tree density (TreeDen) as well as under- tracted samples based on four different criteria (Oksa- story species richness (UnderRich), understory density nen et al., 2011; Appendix S2: Fig. S2). Species coverage (UnderDen), and average understory height (Under- (proportion of the collected species in the Chao-based Height). For more information about the vegetation sur- rarefied diversity) in individual sampling plots was esti- vey see Rembold et al. (2017). mated as 54% Æ 23% in the LS and 72% Æ 24% in the Parameters of litter and soil were assessed in three HE using chao1 in the rareNMtests package in R 3.4.0 subplots (a, b, c): litter pH (CaCl2) was measured using (R Core Team 2017; Appendix S2: Table S1). There were a digital pH meter, litter nitrogen (N) concentration was six sampling plots without identified species (i.e., adult Article e02957; page 4 ANTON M. POTAPOV ET AL. Ecology, Vol. 101, No. 3

FIG. 1. Overview of the four land-use systems and parameters of litter, vegetation, and spider communities investigated. Soil-as- sociated spiders were represented predominantly by the small cryptic species (e.g., Anapidae, lower spider), while LS spiders were represented predominantly by the larger mobile species (e.g., Salticidae, upper spider). Changes in all parameters from rainforest to oil palm monoculture are given in percent. Photo credits of jungle rubber, Jochen Drescher. specimens) in the HE collection and one sampling plot Calculation of energy flux to spider communities without any spiders; there was one sampling plot with- out identified species in the LS collection. Samples with To calculate the flux of energy from prey to spiders, as zero data were included in all analyses and calculations. a proxy of community predation, we first estimated fresh Imaging of representative specimens for each species was body mass for each individual by measuring the body achieved using a custom-built BK Plus Lab System (Dun, length (excluding appendages) and used it in a set of Inc., USA) with integrated Canon EOS (Canon, Ota, body-length–body-mass power equations (Hofer€ and Tokyo, Japan) camera, macro lenses (65 mm and Ott 2009). Second, we calculated metabolic rates from 100 mm), the stacking software Zerene Stacker (Zerene individual body masses (Ehnes et al. 2011) and then cor- Systems, Richland, WA, USA) and formatted in Adobe rected them for assimilation efficiencies during transfer Photoshop CS6 (Adobe Systems, San Jose, CA, USA). All of energy between trophic levels (Barnes et al. 2014, species and images are deposited in the Ecotaxonomy data- 2018). Total flux to spider communities (i.e., predation) base (available online).15 was calculated by summing up individual assimilation- corrected metabolic rates and expressed in kilograms 15 http://ecotaxonomy.org fresh mass per hectare per year by assuming 1 kg fresh March 2020 PREDATION LOSS IN TROPICAL PLANTATIONS Article e02957; page 5 mass = 7 9 106 J (Peters 1983). Calculations included 5.. Body mass, reflecting the ability to handle prey of juvenile spiders; more details on body mass and energy different size (Brose et al. 2019), the resource flux calculations are given in Appendix S3. requirements of the organism (Turney and Buddle 2019), and also dispersal ability, which is important for being able to colonize systems after land-use Assignment of traits and functional indices change. Five size classes based on log10-scaled fresh In order to track differences in functional diversity body mass were distinguished: <0.1 mg, 0.1–1 mg, and functional composition between spider communi- 1–10 mg, 10–100 mg, and >100 mg. ties in forest vs. other land-use systems, we choose functional traits (Violle et al. 2007) that are of high Five trait dimensions well separated the morphos- relevance for hunting and survival, and that could be pecies, in total 60 trait combinations were found inferred from the collected material considering poor (Appendix S1: Fig. S1). The matrix of species’ traits knowledge on the biology of Indonesian spiders. Traits was transformed into a dendrogram using first daisy were assigned at the species level to each of the 139 (Gower’s distances for nominal variables with log- collected spider species; since many species were identi- transformed body size taken as a numeric variable) and fied as morphospecies (i.e., potentially undescribed then hclust in R. Dendrogram was used to calculate species; Appendix S1: Table S2), some trait categories functional diversity (FD) in each spider community were extrapolated from corresponding families. Five (i.e., plot) using treediv in the vegan package. Func- trait dimensions were distinguished, each reflecting dif- tional diversity in this case was calculated as the total ferent aspects of spider predation (effect traits), and branch length among species present in the local com- also responses to environmental/ biotic alterations (re- munity on the trait dendrogram (Petchey and Gaston sponse traits): 2002).

1) Hunting strategy, reflecting how spiders hunt prey. Statistical analysis Six categories were distinguished (Wise 1995, Jocque and Dippenaar-Schoeman 2006, Herberstein 2011): All calculations were done in R 3.5.3 with R studio free hunters, ant mimickers, and four types of web interface 1.0.143 (RStudio, Inc., Boston, MA, USA). builders (orb-type, sheet-type, cob-type and silk-line Since LS and HE sampling events were performed in dif- burrow). Spiders with different hunting strategies ferent years, utilized different sampling areas and cov- vary in the success of catching prey differing in ered different parts of spider community (Appendix S2), behavior and mobility (Ludwig et al. 2018). we present the two sampling methods separately along 2) Body coloration, reflecting the importance of visual with the combined data set. In the analyses, sampling characters in communication with the biotic environ- plots were treated as replicates (N = 32). Effects of land ment. Since coloration often functions to improve use (F, J, R, O) and region (Harapan or Bukit Duabelas) camouflage (Oxford and Gillespie 1998), spiders with were evaluated in a series of generalized linear models different coloration are likely to be confined to cer- (glm); model choice followed the distribution of response tain microhabitats. Five categories were distin- variables and visual inspection of residuals. We first used guished: patterned coloration (clear pattern of spots Poisson distribution for the count data and Gaussian or stripes on the body), dark coloration (clearly pro- distribution for other data. Data on understory density nounced, but uniform coloration), and pale col- were log10-transformed prior to analysis, which oration with full eyes, reduced eyes, and absent eyes. improved homogeneity of variance. Among the parame- 3) Habitat, representing the general spatial niche of spe- ters of litter and vegetation, earthworm abundance, tree cies, primarily corresponding to vertical stratifica- species richness, tree density, and understory species tion. Five categories were distinguished (Jocque and richness were analyzed as count data. Effects of land-use Dippenaar-Schoeman 2006): arboreal (living on system and region on spider abundance (number of col- aboveground parts of plants), arbustive (living on lected individuals per plot, including juveniles), species vegetation close to the ground), ground (living on richness (number of collected species per plot), and num- surface of the litter), cryptic (living within the litter/ ber of species with specific traits (count data) were also soil or using other shelters), and generalist (highly analyzed as count data. We tested for overdispersion in mobile species that are not restricted to a single habi- all Poisson models with dispersiontest in the AER pack- tat). age. In all cases except for species richness of HE spiders, 4) Desiccation resistance, reflecting the ability to remain the models were overdispersed and so we remodeled active in dry habitats or seasons mediated by abdomi- overdispersed data on a negative binomial distribution nal scuta (Main 1976). Since the species with scuta (glm.nb). Effects of land-use system on energy flux to are likely able to endure dryer conditions, we pre- spider communities (including juveniles) and FD were sume that they may hunt in dry periods or in drier analyzed using a Gaussian distribution. Differences microhabitats. Two categories were distinguished: between means were inspected using post hoc Tukey’s presence or absence of scuta. HSD test for multiple comparisons with glth in the Article e02957; page 6 ANTON M. POTAPOV ET AL. Ecology, Vol. 101, No. 3 multcomp package (Hothorn et al. 2008). The model rainforest to the respective land-use system. Significance results are given in Appendix S1: Tables S3–S5. was evaluated using Anova in the car package. Further, we studied how parameters of litter and vege- Due to low number or absence of species collected by tation affect parameters of spider communities indepen- HE in some plots in oil palm monocultures, community dent of land-use system and region. Due to a high composition of HE spiders could not be analyzed sepa- number of parameters to avoid multicollinearity, we rately from LS spiders and we combined both methods. grouped all parameters in (1) habitat structure parame- Since density estimations differed strongly between the ters (TreeDen, UnderDen, UnderHeight, and amount of two sampling methods (see Appendix S2), we used pro- litter), (2) plant diversity parameters (TreeRich and portions of species in total number of identified individ- UnderRich), (3) other soil and litter parameters (litter uals for each community instead of using raw count data pH, litter N concentration, and earthworm abundance). or density estimations. A community matrix with pro- Further, principal component analysis was applied using portions of species or traits, averaged for the two meth- prcomp to each group separately. For habitat structure ods, was used in nonmetric multidimensional scaling parameters PC1 explained 73.5% of the variation and (NMDS) using metaMDS based on Bray-Curtis dissimi- was positively correlated with UnderDen and negatively larities in the package vegan (Oksanen et al., 2011) to with other factors, PC2 explained 12.9% of the variation illustrate differences in species and trait compositions and was positively correlated with UnderDen and among land-use systems. Two dimensions were used for UnderHeight, PC3 explained 9.8% of the variation and NMDS, plots were treated as replicates, and only species was positively correlated with TreeDen and negatively present on more than one plot were analyzed. For the with amount of litter. For plant diversity parameters, trait composition analysis, we used counts of specific PC1 explained 97.0% of the variation and was negatively trait categories (see above); spiders with fresh body mass correlated with TreeRich and UnderRich (Appendix S1: >50 mg were classified as “large” while those with fresh Tables S6, S7). Other soil and litter parameters were body mass <0.1 mg were classified as “small.” Propor- poorly intercorrelated so we did not use principal com- tions of traits were calculated using the number of iden- ponents for them. The most parsimonious model tified individuals per plot as total. Litter and vegetation explaining the abundance, species richness, FD and parameters were fitted in the scaling space to illustrate energy flux of spider communities was selected by com- their effect on the distribution of traits and species; data paring a set of models using AIC (Akaike Information on understory density were log10-transformed prior to Criterion). Due to collinearity of habitat structure and the analysis. Using binary data or proportion of species plant diversity, first these two groups of factors (i.e., in the overall maximum instead of proportions of species their principal components) were compared. Further, a in the total for each community changed the scale, but more parsimonious model was complemented by either showed similar dissimilarity between land-use systems pH, litter N concentration, earthworm density, or all with the same factors being significant (data not shown). these factors together, and the final model was Differences between land-use systems were tested using selected. The model results are given in Appendix S1: analysis of similarities based on species/trait proportions Tables S8, S9. (anosim in package vegan). Data are presented as To compare the rate of species and trait change, we means Æ SD, with P < 0.05 taken as the level of signifi- calculated dissimilarity of species and trait combinations cance. among four land-use systems (plots in each system were merged together) using beta.multi in the betapart pack- RESULTS age. The function estimates overall beta diversity (Soren- sen dissimilarity) and partitions it into nestedness (loss Spider density, energy flux, and diversity of species or trait combination) and turnover compo- nents (replacement of species or trait combinations; Despite a lower number of sampled species and indi- Baselga 2010). To test the difference in the rate of species viduals, the density of HE spiders was one order of mag- and trait change, we analyzed number of species and nitude higher than the density of LS spiders (Fig. 1; number of trait combinations in a single variable number Appendix S2). Spider density and energy flux per unit of forest species/traits. To do that, we selected only spe- area for spiders collected by both HE and LS were gen- cies and trait combinations that were found in rainforest. erally reduced in plantations as compared to rainforest Thereafter, effects of interactions between land-use sys- and this was most prominent in oil palm monoculture tem and parameter type (species or trait combinations) (Fig. 1; Appendix S1: Table S10). The reduction was on the number of forest species/traits were evaluated much more pronounced for the density of HE spiders, using generalized linear models (Poisson distribution). which decreased by 90% from rainforest to oil palm In this case, a significant interaction between the effect monoculture. Compared to rainforest of parameter type and respective land-use system indi- (112 Æ 59 kgÁhaÀ1ÁyrÀ1), energy flux in the spider com- cates a different magnitude in the change in rainforest munity collected by HE was reduced in jungle rubber by species and rainforest trait combinations and, thus, sig- 72%, in rubber monoculture by 55%, and in oil palm nificant differences in species and trait loss/gain from monoculture by 94%. Energy flux in LS spiders was March 2020 PREDATION LOSS IN TROPICAL PLANTATIONS Article e02957; page 7

20 Æ 7kgÁhaÀ1ÁyrÀ1 in rainforest and was reduced by Turnover and loss of species and traits 57% only in oil palm monoculture. In contrast to HE spiders, energy flux in the LS spider communities was Dissimilarity of species among land-use systems (71% similar in rainforest, jungle rubber, and rubber, whereas for HE, 82% for LS, and 78% for HE and LS combined) density was reduced by 30% and 55% in rubber and oil was higher than dissimilarity of trait combinations (64% palm monocultures, respectively, as compared to rain- for HE, 60% for LS, and 57% for HE and LS combined). forest (Fig. 1; Appendix S1: Table S10). For LS spiders and the combined data set, the turnover Spider species richness and functional diversity per component of dissimilarity was higher for species (79% plot were significantly reduced in oil palm monoculture and 74%, respectively) than for trait combinations (54% as compared to rainforest. The loss in species richness, and 49%, respectively). By contrast, these community rarefied richness, and FD per unit area in HE spiders attributes were more similar for HE spiders (52% for was about 75–80% in jungle rubber and rubber mono- species and 48% for traits). The nestedness component culture, and about 91–98% in oil palm monoculture. The of dissimilarity in LS spiders and combined data set loss in these parameters in LS spiders was small in jungle overall was small, but higher for trait combinations than rubber and rubber monoculture, but was 57–63% in oil for species. Nestedness in HE spiders was 19% for spe- palm monoculture (Fig. 2; Appendix S1: Table S10). cies and 15% for trait combinations (Fig. 3). The overall loss in species richness when both sampling Species and functional trait loss occurred at similar methods were combined comprised about 30% in jungle rates in HE spider communities (Fig. 3b). In oil palm rubber and rubber monoculture and 67% in oil palm plantations, no rainforest species and functional trait monoculture as compared to rainforest. Corresponding combinations were found and they were barely replaced losses in FD were slightly less pronounced (20% and by new species/combinations. In LS spider communities, 60%, respectively; Fig. 2). Functional redundancy (esti- species loss was significantly stronger than functional v2 = mated as the ratio between species richness and FD) was trait loss in rubber and oil palm monoculture ( 3 14.8, similar in HE spiders and in LS spiders in rainforest P = 0.0020; Appendix S1: Table S11; Fig. 3c). Despite (1.76 Æ 0.33 and 1.68 Æ 0.32, respectively). Species very few rainforest species appearing in plantations, the richness and FD were intercorrelated across land-use loss of functional trait combinations was less pro- systems, but the slope was steeper in plantation systems, nounced due to other colonizing species with similar suggesting lower functional redundancy when compared functional trait combinations. to rainforest communities. This was consistent in spider Species composition in the combined HE and LS data communities collected by HE, LS, and both methods set differed clearly between more natural (rainforest and combined (Fig. 2). jungle rubber) and more intensified systems (rubber and All studied parameters of vegetation and soil except oil palm monocultures; anosim R = 0.42, P < 0.001). for understory density, pH and earthworm density were The difference in taxonomic composition was visible at negatively affected by the land-use change from rain- the level of families (see Appendix S4: Fig. S1) and forest to monocultures. Understory density was five times clearly pronounced at the level of species (Fig. 4a); 26 higher in oil palm monoculture than in rainforest, while species were found only in rainforest, while 82 species pH and earthworm abundance were increased by 29% were found only in plantation systems. Most of the mea- and 67%, respectively (Fig. 1; Appendix S1: Table S10). sured environmental factors contributed to distinction The most parsimonious model explaining variation in between rainforest/jungle rubber and monocultures, abundance, species richness, and FD for both HE and while differences within these groups were not related to LS spiders included only principle components of habi- measured factors (Fig. 4a). tat structure. In all cases, these spider community Trait composition differed much less than species parameters were negatively related to PC1 (i.e., a positive composition among land-use systems (anosim R = 0.17, effect of the amount of litter, tree density, and under- P = 0.015). Monocultures were characterized by a story height) and to PC3 (i.e., a positive effect of the higher number of generalists, species with patterned col- amount of litter, but a negative effect of tree density), oration, and cob-web building species, while rainforest but were positively related to PC2 (i.e., a positive effect was characterized by species of small size, orb-web of understory density and height). The effects were simi- builders, species covered with scutum, and species living lar in magnitude, but the effect of PC1 was more statisti- in cryptic habitats (Appendix S4: Table S1). Difference cally clear (lower P values) in the majority of cases. The in trait composition was related primarily to the amount effect of PC4 was weak and not significant. Abundance of litter, understory density and height, and pH of HE spiders was also negatively affected by earthworm (Fig. 4b). abundance, although this effect was small. In contrast to abundance and diversity, predation (energy flux) of HE DISCUSSION and LS spiders was better explained by PC1 of plant diversity (a positive effect of tree and understory species We found declines in abundance and biomass and richness), but the effect was not significant for LS spi- strong shifts in the composition of species and func- ders (Appendix S1: Tables S8, S9). tional traits of spider communities with land-use change Article e02957; page 8 ANTON M. POTAPOV ET AL. Ecology, Vol. 101, No. 3

FIG. 2. Species richness and functional diversity in spider communities across the studied land-use systems. In rows: (a) HE and LS spiders combined, (b) HE spiders, and (c) LS spiders. In columns, from left to the right: (1) correlation between species richness and FD values (functional diversity as the total branch length on dendrogram) in spider communities across the studied land-use systems; (2) number of species in each land-use system, means and 95% confidence intervals; (3) FD values in each land-use system, means and 95% confidence intervals. Each point represents a plot, land-use types are shown in different color: rainforest (F, green), jungle rubber (J, gray), rubber monoculture (R, blue), and oil palm monoculture (O, yellow). FD of HE spiders in oil palm mono- culture is not shown since communities comprised only one, if any, identified species in most of the cases. from rainforest to oil palm monoculture. Land-use The differences between the two collection methods change was associated with an almost complete species emphasize that small cryptic spiders, collected by HE, loss and turnover in the spider communities. However, already respond to moderate habitat changes such as the species turnover did not compensate for the loss of conversion of rainforest into jungle rubber, while larger many functional trait combinations. Community energy LS spiders retain higher diversity and biomass in rain- flux was reduced by up to 94% in spiders collected by forest, jungle rubber, and rubber monoculture. Turnover heat extraction (HE) and up to 57% in spiders collected of species was attributed to the loss of small species, orb- by litter sieving (LS) in monoculture plantations, sug- web builders, species with scutum, and species living in gesting a strong reduction in predation pressure by cryptic habitats, and the gain of generalists, species with ground-associated spiders in comparison to rainforests. patterned coloration, and cob-web builders. The stron- These changes were explained the best by habitat struc- ger decline in the small species was surprising as larger ture degradation and change and occurred irrespectively species are often found to react more strongly to distur- of the collection method. bances and land-use change (Brose et al. 2017). March 2020 PREDATION LOSS IN TROPICAL PLANTATIONS Article e02957; page 9

FIG. 3. Spider trait and species turnover with land-use change. In rows: (a) HE and LS spiders combined, (b) HE spiders, and (c) LS spiders. In columns, from left to the right: (1) nestedness and turnover components of Sorensen dissimilarity of species and trait combinations among all land-use systems; (2) changes in the number of species present or absent in rainforest, lines connect mean values; (3) changes in the number of trait combinations present or absent in rainforest, lines connect mean values. Each point represents a plot, land-use types are shown in different color: rainforest (F, green), jungle rubber (J, gray), rubber monoculture (R, blue) and oil palm monoculture (O, yellow). Loss of forest trait combinations is significantly slower than loss of forest species in LS spider communities (Appendix S1: Table S8).

dwelling in the litter and upper soil layer (hereafter “soil- Predation loss in litter and soil associated”). LS favored catches of larger mobile species The magnitude of land-use-driven community change of various families living mainly in the litter, on the litter differed strongly between soil-associated and litter-asso- surface and on understory vegetation above the litter ciated spider communities. Different sampling methods (hereafter “litter-associated”). often provide information on different aspects of tropi- The density and total predation per unit area (energy cal litter and soil invertebrate communities (Sak- flux) of soil-associated spiders exceeded that of litter-as- choowong et al. 2007), and thus a combination of sociated spiders by up to an order of magnitude. More- methods provides better understanding of community over, the data suggest that, except in rainforest, the changes. As the samples of both methods were taken at density of spiders in the top 5 cm of soil markedly the same time of the year, the very pronounced differ- exceeds that in litter (Appendix S2), suggesting that ences in density and taxonomic composition presumably tropical lowland ecosystems are characterized by a large reflect that the two methods represent very different number of small cryptic soil spiders. In part, this was fractions of spider communities, thereby being comple- due to a high incidence of juveniles among soil-associ- mentary (Appendix S2). HE favored catches of small ated spiders (64% of total), but the fraction of juveniles cryptic species, predominantly of the family Oonopidae, in litter-associated spiders was only slightly lower (52%). Article e02957; page 10 ANTON M. POTAPOV ET AL. Ecology, Vol. 101, No. 3

FIG. 4. Differences in species and trait composition between land-use systems illustrated with multidimensional scaling. Land- use types are shown in different color: rainforest (F, green), jungle rubber (J, gray), rubber monoculture (R, blue), and oil palm monoculture (O, yellow). Each point represents one plot; text in panel b represents traits. Combined data of HE and LS are pre- sented. Only species present on more than one plot were analyzed. Litter and vegetation parameters that statistically clear affected the distribution of traits and species are shown with black arrows and labels; other included parameters are shown in gray. Parame- ters are tree species richness (TreeRich), tree density (TreeDen), understory species richness (UnderRich), understory density (UnderDen), average understory height (UnderHeight), amount of litter (Litter), litter nitrogen concentration (Litter N), abun- dance of earthworms (Earthworms), and pH.

Although only few spider species were extracted exclu- most promising approach to maintain natural biological sively from soil, the results suggest that total density and control provided by functionally diverse spider commu- predation by spiders is strongly underestimated if soil- nities. As plant diversity increase structural complexity dwelling species are ignored. in plantations, maximizing both may be possible with lit- In rainforest, the overall predation by soil-associated tle additional efforts (Zemp et al. 2019). spiders exceeded that by litter-associated spiders by a Heat extraction allowed us to show the strong reduc- factor of five. By contrast, the overall predation rates tion in density and energy flux of the small spider species were similar between the two methods in oil palm mono- in jungle rubber, i.e., agroforest systems of moderate culture due to the strong decline in abundance of soil-as- land-use intensity. As suggested by our model, the sociated spiders. Despite the different responses, the decline in density of soil-associated spiders might in part decline in both groups was primarily related to the be related to invasion by earthworms, with their biomass changes in habitat structure, such as understory density being 60–100 times higher in plantation systems and jun- and height and reduction of the litter layer in oil palm gle rubber as compared to rainforest (Potapov et al. monoculture. The effect was stronger for small cryptic 2019). Earthworms may directly consume small soil soil-associated spiders, suggesting that they predomi- invertebrates including hatchlings and eggs of spiders nantly suffer from the loss of litter shelter. A positive (Curry and Schmidt 2007) or disrupt their webs. By con- effect of understory density and height on density and trast, the moderate effect of land use on the large litter- diversity of soil-associated and litter-associated spiders dwelling and arbustive litter-associated spider species supports other studies showing that complex ground suggests that these spiders are less tightly linked to small vegetation promotes ground invertebrate communities in soil prey or more easily switch to alternative prey such plantations (Ashton-Butt et al. 2018). Unlike taxonomic as aboveground invertebrates (Scheu 2001). The and functional diversity, predation by spiders was observed shift in the relative importance of the flux of affected rather by plant diversity than habitat structure. energy through soil-associated and litter-associated spi- These results suggest that enhancing both diversity and der communities suggests principally different pathways habitat structure in plantation monocultures may be the of energy flux through food webs of rainforest and March 2020 PREDATION LOSS IN TROPICAL PLANTATIONS Article e02957; page 11 agricultural land-use systems. Combined data suggest sampling sites that reported major community shifts that land-use systems such as jungle rubber and rubber from rainforest to plantation systems (Schneider et al. monoculture sustain a high functional diversity of 2015, Krashevska et al. 2016, Rembold et al. 2017). The ground spiders, but negatively affect predation of small turnover component (replacement of species) overall cryptic species. By contrast, oil palm monoculture sus- strongly exceeded nestedness of communities across tains only limited functional diversity and predation by land-use systems. Nestedness was higher in soil-associ- spider communities. ated communities, which, together with a strong decline in diversity, suggests that land-use change from rain- forest to monocultures does not open new ecological Rate of species and functional trait loss niches for small cryptic species. Turnover was the main The overall decline in species richness and functional pattern in litter-associated spider communities. We diversity of spider communities in this study is in accor- found 82 species exclusively in plantation systems, sug- dance with observations on the entire litter and soil gesting a large potential of either external colonization invertebrate communities (Barnes et al. 2014, Mumme (dispersal from other habitats to deforested area) or “in- et al. 2015, Potapov et al. 2019). The two sampling meth- ternal resilience” (reproduction of rare species, that were ods used in our study demonstrated two different aspects present in the community before deforestation). of community responses to land-use change. The soil-as- Trait composition of spider communities was affected sociated spider community responded much stronger by habitat-structural parameters such as density of and the loss of species occurred at the same rate as the understory, understory plant height, and the amount of loss of functional trait combinations. Density, biomass litter. This may reflect a strong link between functional and energy flux per unit area declined by more than composition of arthropod community and the structural 90%. These strong changes led to a simultaneous loss of complexity and structural characteristics of habitats sets of redundant species and new communities assem- (Lassau et al. 2005). In particular, the amount of litter in bled including few species with new functional trait com- rubber, and especially in oil palm plantations, is lower binations. By contrast, changes in the litter-associated than in rainforest (Krashevska et al. 2015). Species of spider community were less pronounced with the loss of the families Oonopidae and Tetrablemmidae; i.e., small species exceeding the loss of functional trait combina- spiders using leaf litter as shelter and for hunting tions; 10–17% of forest species, but 31–48% of forest (Jocque and Dippenaar-Schoeman 2006), suffered the trait combinations, were found in monoculture planta- most (Appendix S4). The absence of litter in oil palm tions. We also found a large overlap in trait, but not spe- plantations also affects the composition of other ground cies composition, between land-use systems. The arthropods such as beetles (Chung et al. 2000). The stronger reduction in species richness as compared to reduction in leaf litter may also explain the absence of functional group richness is in line with that of the entire spiders with reduced vision, and the reduction in litter invertebrate community (Mumme et al. 2015). ground, cryptic and small-sized species dwelling in shel- Since functional redundancy was similar in soil-associ- ters. The increase in understory plant density may ated and litter-associated spiders in rainforest, it is likely explain the presence of arbustive species and species that the loss of species and functional traits with rain- implying three-dimensional cob webs. Patterned col- forest conversion was related to the magnitude of distur- oration occurred more frequently in monoculture plan- bance (land-use change more strongly affected soil- tations, suggesting that in more open habitats visual associated spiders). The litter-associated spider commu- cues are more important and favor mobile, visual-hunt- nity was only moderately affected by land-use change ing, predator species. Thus, the structural components and the overall set of functional trait combinations of habitat are likely to support the diversity and shape changed slowly due to colonization by new species with the composition of functional traits in communities of similar functional traits as in rainforest species, but pre- ground-associated invertebrate predators. sumably coping better with the new environment than rainforest species (Appendix S4: Fig. S2). As a result, CONCLUSIONS overall species richness and functional diversity changed little, except in oil palm monoculture. Our study is the first to use a trait-based approach for analyzing tropical spider communities to understand the drivers of community turnover and invertebrate preda- Changes in community composition tion loss with land-use change. Conversion of rainforest Land-use change from rainforest to monoculture to monoculture systems strongly affected ground spider plantation systems was associated with an almost com- communities, resulting in the loss of diversity and ecosys- plete species loss and turnover in the spider communities tem functions they provide; total predation, estimated as with <20% of rainforest species being present in rubber energy flux to spiders, was reduced by 94% and 57% in and oil palm monocultures, irrespective of the collection oil palm monoculture for both soil-associated and litter- method. This is in line with previous studies on prokary- associated spider communities, respectively. Land-use otes, testate amoebae, and plants from the same systems such as jungle rubber and rubber monocultures Article e02957; page 12 ANTON M. POTAPOV ET AL. Ecology, Vol. 101, No. 3 sustain a high functional diversity of mobile litter-dwell- non-native species as pollinators of endemic Hawaiian plants: ing and arbustive spiders, but negatively affect predation non-native pollinators. Conservation Biology 28:478–488. of small cryptic species. Oil palm monocultures sustain Bardgett, R. D., and W. H. van der Putten. 2014. Belowground biodiversity and ecosystem functioning. Nature 515:505–511. only limited functional diversity and predation by spider Barnes, A. D., M. Jochum, S. Mumme, N. F. Haneda, A. Fara- communities. Small cryptic spiders, which are often jallah, T. H. Widarto, and U. Brose. 2014. Consequences of ignored in ecological studies, form an important compo- tropical land use for multitrophic biodiversity and ecosystem nent of spider communities in tropical rainforests, but are functioning. Nature Communications 5:5351. less abundant in plantation systems. Barnes, A. D., M. Jochum, J. S. Lefcheck, N. Eisenhauer, C. Plantation systems are colonized by many spider spe- Scherber, M. I. O'Connor, P. de Ruiter, and U. Brose. 2018. cies, in part replacing species being excluded, but this Energy flux: the link between multitrophic biodiversity and ecosystem functioning. Trends in Ecology & Evolution turnover does not compensate the losses in functional 33:186–197. diversity and predation per unit area. The ratio of spe- Barrion, A. T., J. A. Litsinger, C. A. B. International, and I. R. cies-to-trait losses and loss of functions likely depend on R. Institute.1995. Riceland spiders of south and Southeast the magnitude of disturbance, with moderate changes Asia. CAB International, Wallingford, UK. allowing for colonization of new species that play similar Baselga, A.. 2010. Partitioning the turnover and nestedness functional roles as the species being lost, which gives components of beta diversity: partitioning beta diversity. Glo- bal Ecology and Biogeography 19:134–143. some promise for potential conservation practices. Func- Brose, U., et al. 2017. Predicting the consequences of species tional diversity and composition were primarily related loss using size-structured biodiversity approaches: Conse- to the structural complexity of the habitat, specifically to quences of biodiversity loss. Biological Reviews 92:684–697. the amount of litter and ground vegetation density and Brose, U., et al. 2019. Predator traits determine food-web archi- height, while predation by spiders was related to plant tecture across ecosystems. Nature Ecology & Evolution – diversity. Thus, the negative impacts of land-use change 3:919 927. Chung, A. Y. C., P. Eggleton, M. R. Speight, P. M. Hammond, on spider communities might be mitigated by adopting and V. K. Chey. 2000. The diversity of beetle assemblages in management strategies such as mulching and reduced different habitat types in Sabah, Malaysia. Bulletin of Ento- weeding/herbicide application (Ashton-Butt et al. 2018, mological Research 90:475–496. Tao et al. 2018) or intercropping monocultures with Clough, Y., et al. 2016. Land-use choices follow profitability at other trees (Teuscher et al. 2016). Further, to improve the expense of ecological functions in Indonesian smallholder colonization of monoculture plantations from refuge landscapes. Nature Communications 7:13137. areas, our findings suggest that plantations are best inte- Corley, R. H. V., and P. B. Tinker. 2016. The oil palm. Fifth edi- tion. Wiley-Blackwell, Hoboken, New Jersey, USA grated into complex landscapes comprising different Curry, J. P., and O. Schmidt. 2007. The feeding ecology of earth- habitat types including rainforest. worms—a review. Pedobiologia 50:463–477. Deeleman-Reinhold, C. L. 2001. Forest spiders of South East ACKNOWLEDGMENTS Asia: with a revision of the sac and ground spiders (Araneae: This study was funded by the Deutsche Forschungsgemein- Clubionidae, Corinnidae, Liocranidae, Gnaphosidae, Pro- schaft (DFG), project number 192626868–SFB 990 in the frame- didomidae and Trochanterriidae). Brill, Leiden, The Nether- work of the collaborative German-Indonesian research project lands. CRC990. We thank the following persons and organizations for Dislich, C., et al. 2017. A review of the ecosystem functions in granting us access to and use of their properties and scientific oil palm plantations, using forests as a reference system. Bio- – cooperation during the field work: local plot owners, PT REKI, logical Reviews 92:1539 1569. Bukit Duabelas National Park, Achmad Farajallah, Noor Fari- Drescher, J., et al. 2016. Ecological and socio-economic func- khah Haneda, Tri Heru Widarto, Sri Sudarmiyati Tjitrosoe- tions across tropical land use systems after rainforest conver- dirdjo, and Bambang Hariyadi. This study was conducted using sion. Philosophical Transactions of the Royal Society B organisms collected based on Collection Permits No. S.07/KKH- 371:20150275. 2/2013, 2695/IPH.1/KS.02/XI/2012, 25/EXT/SIP/FRP/SM/III/ Ehnes, R. B., B. C. Rall, and U. Brose. 2011. Phylogenetic 2013, 17/EXT/SIP/FRP/SM/III/2014, and 1313/FRP/SM/VIII/ grouping, curvature and metabolic scaling in terrestrial inver- – 2015 recommended by the Indonesian Institute of Sciences tebrates: invertebrate metabolism. Ecology Letters 14:993 (LIPI) and issued by the Ministry of Forestry (PHKA). 1000. FAOSTAT. 2016. FAOSTAT Statistical database. http://www.fa o.org LITERATURE CITED Fayle, T. M., E. C. Turner, J. L. Snaddon, V. K. Chey, A. Y. C. Chung, P. Eggleton, and W. A. Foster. 2010. Oil palm expan- Allen, K., M. D. Corre, S. Kurniawan, S. R. Utami, and E. sion into rain forest greatly reduces ant biodiversity in Veldkamp. 2016. Spatial variability surpasses land-use change canopy, epiphytes and leaf-litter. Basic and Applied Ecology effects on soil biochemical properties of converted lowland – – 11:337 345. landscapes in Sumatra, Indonesia. Geoderma 284:42 50. Fierer, N., M. S. Strickland, D. Liptzin, M. A. Bradford, and C. Ashton-Butt, A., et al. 2018. Understory vegetation in oil palm C. Cleveland. 2009. Global patterns in belowground commu- plantations benefits soil biodiversity and decomposition rates. nities. Ecology Letters 12:1238–1249. Frontiers in Forests and Global Change 1. https://doi.org/10. Fitzherbert, E. B., M. J. Struebig, A. Morel, F. Danielsen, C. A. 3389/ffgc.2018.00010 Bruhl,€ P. F. Donald, and B. Phalan. 2008. How will oil palm Aslan, C. E., E. S. Zavaleta, B. Tershy, D. Croll, and R. H. expansion affect biodiversity? Trends in Ecology & Evolution Robichaux. 2014. Imperfect replacement of native species by 23:538–545. March 2020 PREDATION LOSS IN TROPICAL PLANTATIONS Article e02957; page 13

Flynn, D. F. B., M. Gogol-Prokurat, T. Nogeire, N. Molinari, Larsen, T. H., N. M. Williams, and C. Kremen. 2005. Extinc- B. T. Richers, B. B. Lin, N. Simpson, M. M. Mayfield, and F. tion order and altered community structure rapidly disrupt DeClerck. 2009. Loss of functional diversity under land use ecosystem functioning. Ecology Letters 8:538–547. intensification across multiple taxa. Ecology Letters 12:22– Lassau, S. A., D. F. Hochuli, G. Cassis, and C. A. M. Reid. 33. 2005. Effects of habitat complexity on forest beetle diversity: Foster, W. A., et al. 2011. Establishing the evidence base for do functional groups respond consistently? Diversity and maintaining biodiversity and ecosystem function in the oil Distributions 11:73–82. palm landscapes of South East Asia. Philosophical Transac- Ludwig, L., M. A. Barbour, J. Guevara, L. Aviles, and A. L. tions of the Royal Society B 366:3277–3291. Gonzalez. 2018. Caught in the web: Spider web architecture Gagic, V., et al. 2015. Functional identity and diversity of ani- affects prey specialization and spider–prey stoichiometric mals predict ecosystem functioning better than species-based relationships. Ecology and Evolution 8:6449–6462. indices. Proceedings of the Royal Society B 282:20142620– Main, B. Y. 1976. Spiders. Collins, Sydney, New South Wales, 20142620. Australia. Garcıa, D., D. Martınez, D. B. Stouffer, and J. M. Tylianakis. Margono, B. A., P. V. Potapov, S. Turubanova, F. Stolle, and M. 2014. Exotic birds increase generalization and compensate for C. Hansen. 2014. Primary forest cover loss in Indonesia over native bird decline in plant-frugivore assemblages. Journal of 2000–2012. Nature Climate Change 4:730–735. Animal Ecology 83:1441–1450. Mason, N. W. H., D. Mouillot, W. G. Lee, and J. B. Wilson. Georg-August-Universitat€ Gottingen.€ 2019. Collaborative 2005. Functional richness, functional evenness and functional Research Centre 990: Ecological and Socioeconomic Func- divergence: the primary components of functional diversity. tions of Tropical Lowland Rainforest Transformation Sys- Oikos 111:112–118. tems (Sumatra, Indonesia) (Phase 2: 2016 – 2019). http:// Mumme, S., M. Jochum, U. Brose, N. F. Haneda, and A. D. www.uni-goettingen.de/en/310995.html Barnes. 2015. Functional diversity and stability of litter-inver- Gibbs, H. K., A. S. Ruesch, F. Achard, M. K. Clayton, P. tebrate communities following land-use change in Sumatra, Holmgren, N. Ramankutty, and J. A. Foley. 2010. Tropical Indonesia. Biological Conservation 191:750–758. forests were the primary sources of new agricultural land in Nyffeler, M., and K. Birkhofer. 2017. An estimated 400–800 the 1980s and 1990s. Proceedings of the National Academy of million tons of prey are annually killed by the global spider Sciences USA 107:16732–16737. community. Science of Nature 104:30. Gurr, G. M., S. D. Wratten, D. A. Landis, and M. You. 2017. Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, R. B. Habitat management to suppress pest populations: progress O'hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, and H. and prospects. Annual Review of Entomology 62:91–109. Wagner. 2011. Vegan: community ecology package. R pack- Hassall, M., D. T. Jones, S. Taiti, Z. Latipi, S. L. Sutton, and age version 1.17-6. R package version:1–17. https://CRAN.R- M. Mohammed. 2006. Biodiversity and abundance of terres- project.org/package=vegan trial isopods along a gradient of disturbance in Sabah, East Oxford, G. S., and R. G. Gillespie. 1998. Evolution and ecology Malaysia. European Journal of Soil Biology 42:S197–S207. of spider coloration. Annual Review of Entomology 43:619– Herberstein, M. E. 2011. Spider behaviour: flexibility and versa- 643. tility. Cambridge University Press, Cambridge, UK. Petchey, O. L., and K. J. Gaston. 2002. Functional diversity Hofer,€ H., and R. Ott. 2009. Estimating biomass of Neotropical (FD), species richness and community composition. Ecology spiders and other arachnids (Araneae, Opiliones, Pseudoscor- Letters 5:402–411. piones, Ricinulei) by mass-length regressions. Journal of Peters, R. H..1983. The ecological implications of body size. Arachnology 37:160–169. Cambridge University Press, Cambridge, UK. Hothorn, T., F. Bretz, and P. Westfall. 2008. Simultaneous infer- Potapov, A. M., B. Klarner, D. Sandmann, R. Widyastuti, and ence in general parametric models. Biometrical Journal S. Scheu. 2019. Linking size spectrum, energy flux and 50:346–363. trophic multifunctionality in soil food webs of tropical land- Jocque, R., and A. S. Dippenaar-Schoeman. 2006. Spider fami- use systems. Journal of Animal Ecology 88:1845–1859. lies of the world. Musee Royal de l'Afrique Central, Tervuren, Prieto-Benıtez, S., and M. Mendez. 2011. Effects of land man- Belgium. agement on the abundance and richness of spiders (Ara- Kempson, D., M. Lloyd, and R. Ghelardi. 1963. A new extrac- neae): a meta-analysis. Biological Conservation 144:683– tor for woodland litter. Pedobiologia 3:1–21. 691. Klarner, B., H. Winkelmann, V. Krashevska, M. Maraun, R. R Core Team.2017. R: A language and environment for statisti- Widyastuti, and S. Scheu. 2017. Trophic niches, diversity and cal computing. http://www.R-project.org community composition of invertebrate top predators (Chilo- Rembold, K., H. Mangopo, S. S. Tjitrosoedirdjo, and H. Kreft. poda) as affected by conversion of tropical lowland rainforest 2017. Plant diversity, forest dependency, and alien plant inva- in Sumatra (Indonesia). PLoS ONE 12:e0180915. sions in tropical agricultural landscapes. Biological Conserva- Krashevska, V., B. Klarner, R. Widyastuti, M. Maraun, and S. tion 213:234–242. Scheu. 2015. Impact of tropical lowland rainforest conversion Rigal, F., P. Cardoso, J. M. Lobo, K. A. Triantis, R. J. Whit- into rubber and oil palm plantations on soil microbial com- taker, I. R. Amorim, and P. A. V. Borges. 2018. Functional munities. Biology and Fertility of Soils 51:697. traits of indigenous and exotic ground-dwelling arthropods Krashevska, V., B. Klarner, R. Widyastuti, M. Maraun, and S. show contrasting responses to land-use change in an oceanic Scheu. 2016. Changes in structure and functioning of protist island, Terceira, Azores. Diversity and Distributions 24:36– (testate Amoebae) communities due to conversion of lowland 47. rainforest into rubber and oil palm plantations. PLoS ONE Sakchoowong, W., S. Nomura, K. Ogata, and J. Chan- 11:e0160179. paisaeng. 2007. Comparison of extraction efficiency between Laliberte, E., and P. Legendre. 2010. A distance-based frame- Winkler and Tullgren extractors for tropical leaf litter work for measuring functional diversity from multiple traits. macroarthropods. Thai Journal of Agricultural Science Ecology 91:299–305. 40:97–105. Article e02957; page 14 ANTON M. POTAPOV ET AL. Ecology, Vol. 101, No. 3

Scheu, S. 2001. Plants and generalist predators as links between mass in Sabah, Malaysia. Journal of Tropical Ecology the below-ground and above-ground system. Basic and 25:23–30. Applied Ecology 2:3–13. Turney, S., and C. M. Buddle.2019. Body size mediates the rela- Schneider, D., et al. 2015. Impact of lowland rainforest transfor- tionship between spider (Arachnida: Araneae) assemblage mation on diversity and composition of soil prokaryotic com- composition and prey consumption rate: results of a meso- munities in Sumatra (Indonesia). Frontiers in Microbiology cosm experiment in the Yukon, Canada. Oecologia 189:757– 6:1339. 768. Tao, H.-H., J. L. Snaddon, E. M. Slade, L. Henneron, J.-P. Cali- Vandewalle, M., et al. 2010. Functional traits as indicators of man, and K. J. Willis. 2018. Application of oil palm empty biodiversity response to land use changes across ecosystems fruit bunch effects on soil biota and functions: a case study in and organisms. Biodiversity and Conservation 19:2921– Sumatra, Indonesia. Agriculture, ecosystems & environment 2947. 256:105–113. Violle, C., M.-L. Navas, D. Vile, E. Kazakou, C. Fortunel, I. Teuscher, M., et al. 2016. Experimental biodiversity enrichment Hummel, and E. Garnier. 2007. Let the concept of trait be in oil-palm-dominated landscapes in Indonesia. Frontiers in functional! Oikos 116:882–892. Plant Science 7:1538. Wise, D. H..1995. Spiders in ecological webs. Cambridge Tscharntke, T., A. M. Klein, A. Kruess, I. Steffan-Dewenter, University Press, Cambridge, UK. and C. Thies. 2005. Landscape perspectives on agricultural Zemp, D. C., M. Ehbrecht, D. Seidel, C. Ammer, D. Craven, J. intensification and biodiversity–ecosystem service manage- Erkelenz, B. Irawan, L. Sundawati, D. Holscher,€ and H. ment. Ecology Letters 8:857–874. Kreft. 2019. Mixed-species tree plantings enhance structural Turner, E. C., and W. A. Foster. 2009. The impact of forest complexity in oil palm plantations. Agriculture, Ecosystems conversion to oil palm on arthropod abundance and bio- & Environment 283:106564.

SUPPORTING INFORMATION Additional supporting information may be found in the online version of this article at http://onlinelibrary.wiley.com/doi/ 10.1002/ecy.2957/suppinfo

DATA AVAILABILITY Data are available on Figshare: https://doi.org/10.6084/m9.figshare.10305506